At Meta, I built a scalable Python ML inference service with Redis caching, improving p99 by 28%. I also developed a feature-generation pipeline that improved training-data freshness by 2.3×.
At Cerebri AI, I engineered data ingestion and ETL systems, bringing pipeline reliability to 99.5% and eliminating duplicate-record incidents. I also created data-quality checks that increased successful training cycles by 35%.
At Google, I designed event-processing pipelines and tuned analytical queries, reducing average job runtimes by 40%. Earlier, at SparkCognition, I developed predictive models and repeatable data-preprocessing workflows.
